Delhi Technological University: Designand Development of the Littoral AUV

Zyra 2.0

Aayush Gupta,Vatsal Rustagi,Raj Kumar Saini,Akshay Jain,Prateek Murgai,Aditya RastogiPronnoy Goswami,Ayush Tomar,Abhijit Ray,Bhavy Dikshit,Prithvijit Chattopadhyay

Shubham Raina,Mayank Gupta,Vikas Kumar Singh,Shubham Jain,Chirag Gupta

Abstract—The Delhi Technological University Au-tonomous Underwater Vehicle project team’s main ob-jective is to design and develop an autonomous under-water vehicle for the AUVSI and ONR InternationalRobosub-2014 competition. The competition is held at theTRANSDEC facility, part of SPAWAR Systems CenterPacific in San Diego, California. The paper presents thedesign and development of a modular littoral autonomousunderwater vehicle called ’Zyra 2.0’ having four degreesof freedom for performing the following tasks underwatertarget localization and homing, buoy detection, path fol-lowing, obstacle detection and obstacle manipulation tasks.The development of the AUV has been divided into fivedepartments namely mechanical design and fabrication,embedded and power systems, control and software, imageprocessing, underwater acoustics. A fully functional AUVhas been tested in a self created arena with different tasksspread out in a shallow water environment.

I. Introduction

Autonomous Underwater Vehicle are powerfuland complex systems which are capable of perform-ing underwater (shallow and deep sea) tasks. Therange encompasses bathymetry calculation, detec-tion of faults in oil pipelines, recovery and moni-toring of submerged installations and even complextasks like collecting data which aids in understand-ing global warming. The aim is to design anddevelop an Autonomous Underwater Vehicle whichwill serve as a technology demonstrator which ishighly compact, multirole and can be used forvarious missions with the independence of choosingpayload. Zyra 2.0 is the product of completely re-

fined mechanical assembly, control systems, embed-ded systems, vision system and acoustics module.The mechanical model of Zyra 2.0 is designed to bein hydrodynamically stable equilibrium both belowand above water surface. The mechanical systemconsist of main pressure hull, front and back lids,frame, electronics rack. The focus of embeddedand the power systems department is mainly onthe implementation of acoustic positioning module,actuator board and the designing of a power distri-bution board for diverting power to various moduleson the AUV [1]. Dynamic control is achieved byretrieval of coordinates using various PID controlalgorithms [2],[3]. The software is designed to runin decentralized multi-threaded agent architecture,with the threads handling pressure sensor, acoustics,cameras, control system, IMU, each performinginput and output operations in continuous loops.The mission plan model is also taken care bythe control department as it is responsible for theartificial intelligence of the vehicle [4],[5]. Thevision system is designed to perform tasks like buoydetection, bin detection, path detection, entry gatedetection. The image processing software consistsof various colour filters, edge and line detectionalgorithms. For the target(sound source or pinger)localization the AUV employs a three dimensionalarray of hydrophones forming a passive SONARsystem. The data acquired through hydrophonesis passed through a amplifier, filter and detectorcircuit before it is converted into its digital form

for further processing. The localization task is asequential task in which first we calculate a setof values of TDOA( Time Difference of Arrival)using the hydrophone signals and using these valueswe estimate the bearing and the position vector ofthe sound source with respect to the AUV. For theexperimental testing of the AUV an arena is createdin a shallow water area and the tasks of entry gatedetection, buoy detection, path detection and soundsource are spread throughout the arena.

II. Mechanical Design

Fig. 1. Solidworks Assembly of Zyra 2.0

Zyra 2.0 is the product of indigenously designedmechanical assembly to embedded and control sys-tems. It has a custom made LiPo battery pack, anovel power distribution board, redesigned actua-tor control board, battery monitor, leak sensor, anacoustic signal processing module and improvedsoftware. The fabrication material chosen for pres-sure hull is Virgin Cast Acrylic (Clear) because ofits excellent workability, strength, shock resistanceand comparatively low density. Simulation is doneon hull at underwater depth of 10 m and Factor ofsafety is found out to be more than 5.1.

The body shall be propelled by four thrustersand has a net positive buoyancy. The frame isso designed to ensure easy mounting of thrusters,sensors, grabbers etc. The profile of the vehiclemakes it highly manoeuvrable. The front lid is made

Fig. 2. Solidworks simulation of the main hull-Von Mises Stressdistribution

Fig. 3. Exploded view of assembly

of virgin cast acrylic .Its transparency ensures clearand unobstructed view for the forward camera thatis mounted inside the hull.Back-lid will be made ofaluminium.Its smooth surface will ensure easy andefficient installation of connectors and aluminiumbeing a good thermal conductor will also help as aheat sink.

A. Electronics Hull

The main hull of Zyra 2.0 is of cylindrical shape.The shape of the vehicle has been decided aftercalculations, keeping various hydrodynamic param-eters in mind to improve the overall performanceof the robot. Main frame is designed to ensureeasy installation of thrusters, grabber, hydrophonearray etc. The main consideration while designingit was that unwanted stresses on hull are uniformlyand efficiently transmitted to the whole frame andprevent cracks. Some of its components are madeup of Stainless Steel GR316 and others are of

Delhi Technological University Autonomous Underwater Vehicle

Fig. 4. Solidworks model of main frame

Aluminium. The reason for this approach is tohave a suitable combination of weight, buoyancy,resistance to corrosion along with low maintenance,high strength and ease of fabrication. The hull iswaterproofed using Toggle Clamps instead of Rodsfor faster opening and closing of the rear lid.

B. Vehicle DynamicsThrusters: Four strap-on BTD-150 thrusters from

Seabotix Inc. are used to manoeuvre the vehicle.Two thrusters facilitate motion in the horizontaldirection as well as yaw motion when operatedin differential mode, while other two facilitate thevertical motion. These thrusters provide a two bladebollard thrust of 2.9 kgf and require power in therange of 80-110 watts depending on the conditionsoutside. All electronic components are housed in-

Fig. 5. Solidworks model of electronics frame

side the electronics hull. Components are placed inracks inside the main hull so as to facilitate easyremoval of component in case of a technical fault.The rack is so designed to ensure easy additionof new shelves later as per requirements and leastlength of wires. Battery is placed at lower mostshelf to ensure lower centre of gravity and improvestability and resist roll.

C. Degrees OF Freedom

Fig. 6. Zyra 2.0’s Degrees of Freedeom

The four thrusters are installed in such a waythat the vehicle has four degrees of freedom namelyheave, yaw, pitch and surge.

D. Stability

The AUV is designed in such a way so thatthe centre of gravity and centre of buoyancy liein the same vertical line.Fig.7 displays the stabilityparameters.

Fig. 7. Centre of buoyancy and gravity in same vertical line

Delhi Technological University Autonomous Underwater Vehicle

TABLE ICharacterisitics of Zyra 2.0

Property ValueDimensions 70cm x 40cm x 50 cm

Diameter of Hull 30cmDry Weight 35Kg

Length of Hull 37cmPropulsion 2 horizontal + 2 vertical thrusters

Velocity 0-1.5m/s

E. Robustness

Fig.2 displaying Von Mises Stress Distributionand Fig.8 showcase the robustness of the mechanicaldesign.

Fig. 8. Depicts the fixture and face of pressure application

III. Embedded and Power Systems

As discussed before for the AUV Zyra 2.0 thefocus of electronics department is primarily onimplementation of SONAR and signal processingmodule, actuator board for control of four SeaBotixthrusters. A new power distribution board is de-signed for voltage regulation and to provide powerdirectly to all the systems through one particularboard which would eliminate the need of furtherbucking of voltage and hence concentrates the heatdissipating unit in one particular section of hullwhich can be placed near the metal ends to allowheat exchange with the water.

A. Power Systems

The electrical power system is comprised oflithium polymer (LiPo) battery, encased within themain electronics hull. The battery is rated at 18000

mAh at 18.5V. The battery protection board mon-itors voltage of each cell and shuts off the powerif the voltage drops to 16V, also the overchargeand current drawn is monitored by the same. Itis connected to a Power Distribution Board (PDB)which is responsible for diverting power to variousmodules on the bot. The PDB is equipped with fusesin case of battery/circuit failure, multiple capacitorsto smooth out ripples, and eliminates high frequencynoise. The PDB divides the power and providesregulated power supply to sensitive electronics. Itprovides Zyra 2.0 with clean, isolated, and regulatedpower at 5V, 12V, and 18.5V.

B. Actuator ControlThe actuator board used for Zyra 2.0 is cus-

tom designed. It takes in care the electromagneticinterference from the thrusters which distorts themotor control signal coming from the microcon-troller unit LPC-2148(ARM). This microcontrolleris responsible for taking in signals from the centralprocessing board through packetized serial interface.The microcontroller is capable of processing at30MIPS.

Fig. 9. Schematic for Actautor board

The motor controllers from Dimension Engineer-ing are capable of functioning at ultra-high fre-

Delhi Technological University Autonomous Underwater Vehicle

quency of 32 KHz thus making it inaudible forhuman ears and eliminate the irritating hummingnoise. The controllers can be configured to use ei-ther in tank style differential drive or analog voltagecontrol. The microcontroller unit receives signalsspecifying direction and speed of the Zyra 2.0whichin turn actuate the required thrusters. A hermeticallysealed switch is responsible for killing the power tothe propulsion system.A normally open reed switchis responsible for killing power to the propulsionsystem. The advantage is it can be easily operatedfrom outside the main hull.

Along with hard kill a soft kill is also incor-porated in the code. Analog voltage is used tocontrol the speed of thrusters with another signalto specify the direction of motion. PWM motorcontroller of dsPIC30F2010 is used to generatevariable duty cycle PWM signal which is filteredand smoothed into an analog signal (0V-5V) via ahigh capacitance, RC filter. The added advantageof mounting syren-10 on the actuator board is thatany particular controller can be replaced in caseit gets damaged and thus prevailing re-usabilityof the board. Learning from past experiences, the

Fig. 10. Actuator Board

propulsion system of Zyra 2.0 have been completelyredesigned.The communication of Master Computerand Micro-controller have been changed from RS-232 protocol to TTL. Instead of the fact that RS-232 is noise tolerant, the communication to UARTis done using USB-TTL chips.

C. Electronics RackModularity of the vehicle has been the priority

while developing the vehicle. Electronics Rack is

designed such that it holds all the electronics andcan be removed from Zyra 2.0 without disturbingthe rest of the system

Fig. 11. Electronics rack for Zyra 2.0

The level of various racks can be varied accordingto payload.

Fig. 12. Orientation data received from IMU used as an input inControl system module.

IV. MissionModuleMission plan module of Zyra 2.0 is responsible

for the artificial intelligence of the vehicle. It is atthe highest level in the software hierarchy, coordi-nating the global state of the AUV and the state ofeach subsystem. It makes calls to sensor moduleslike vision, sound etc. determine the position and

Delhi Technological University Autonomous Underwater Vehicle

TABLE IICommericial Off the Shelf(COTS) products used in Zyra 2.0

Devices ModelKontron Motherboard 986LCD-M/mITX

Battery Li-Po 18.5V , 18000mAhMotor Controller Syren 10

Camera Microsoft LifeCam-3000 and Logitech HD Pro C920Propulsion SeaBotic BTD-150

Pressure Sensor Applied Measurements Pi9933IMU XSENS MTi-28A

Actuator Controller LPC-2148(ARM)Servo HiTec HS5646WPHydrophones Reson TC-4013

orientation of the AUV and to identify targets inthe arena. The mission plan module coordinates thestate of the AUV as it goes through the entire mis-sion arena. It has a scheduler/ timing module whichtimes each operation and is capable of making smartdecisions of leaving a task and moving to the nextone based on mission time elapsed and pre-writtencontingency plans. Once the AUV determines whattype of action is to be performed, it calls the controlmodule which commands the actuators to functionprecisely.

V. ControlModule

Control module is called by the mission planmodule as and when required to change the orienta-tion and position and orientation of the AUV basedon the operation being performed and input fromvision, acoustic, depth and other sensor modules.Control module maintains the orientation of AUVusing continuous PID loops running simultaneously.It relies on the mechanical stabilization for both rolland pitch movement, and thus, only the yaw, depthand horizontal movement of AUV is controlled bythis module. The PID control algorithm has beencoded in Qt Creator 3.0. This method has proven tobe more efficient, less processor intensive and easilyimplementable. The system attempts to maintainits state using dynamic feedback from the IMU,pressure sensor and the acoustic and vision modules.User interfaces have been specifically developed totune and adjust the PID parameters easily.

This year we have shifted all our codes to opensource softwares. We have integrated all the codesin Qt SDK 5.2 using Qt Creator 3.0.The Computer

Fig. 13. Various software modules working in tandem to achievemission control

Vision module is developed using OpenCV libraryof C++.

The Computer Vision module was developed us-ing the OpenCV library of C++ integrated with QtSDK using Ubuntu 12.04. The high parallelism dur-ing execution of programs on multi-core processorsin Ubuntu (Linux) gives the vision module the re-

Delhi Technological University Autonomous Underwater Vehicle

quired real-time computational power. The moduleincorporates concepts involving the various imageprocessing algorithms for particle analysis, imagesegmentation, binary morphology, Hough transfor-mation and Machine Vision. The major change thisyear is that the navigation system works on absoluteyaw (angle) control and we have used the UserDatagram Protocol (UDP) for communication ofthe AUV. The previous generation vehicles had aless accurate navigation system partly based onheuristics.

We have used an Android phone as our IMU(Inertial Measurement Unit) using an application forAndroid operating System to get the orientation ofthe AUV as an alternative of our IMU. The Androidphone uses the UDP communication for sending andreceiving data packets between the Android phoneand the single board computer.

VI. Software andMachine Vision

The software team is responsible for missionplanning, machine vision and developing severalsoftware tools to enhance the debugging and oper-ability of the vehicle. This year the team has shiftedall the codes to open source softwares keepingin mind the following objectives: 1)Enhancing theparallel processing capability and modularity. 2)Re-ducing mission task execution time. 3)Developmentof debugging and replacement modules in case ofcritical hardware failure.

Fig. 15. Generalized flow chart for target detection using ImageProcessing

A. Software stack on SBCThe software stack has been built on Ubuntu

Core GNU/Linux distribution, using Ubuntu 12.04.The vehicles computer is composed of KontronMotherboard 986LCD-M/mITX, 4 GB of RAM andan Intel Atom Processor clocked at 1.60GHz . Theentire software stack has been written in C/C++.The software stack performs mission planning, im-age processing, and handling the communicationbetween the hardware modules, power board andinertial measurement unit.

B. MiddlewareThe objective of the middleware is to formulate

a neat and clean interface for various processes tointeract with each other. The inter process commu-nication for various processes and multiple sensormodules on the on-board computer operate on theshared memory managed on Qt SDK 5.2 usingQSharedMemory Class. A central mission controlmodule then takes the decisions according to thesensor values received from these processes.

C. VisionThe team has integrated all the codes in Qt SDK

5.2 using Qt Creator 3.0. The Computer Visionmodule is developed using OpenCV library of C++.We are using two cameras (Logitech C920) onefacing downwards and the other forward facingcamera. These cameras are USB 2.0 enabled andinterfaced to the SBC. The front facing camera isused for localization while the bottom camera isused for orienting the vehicle along the path placedat the bottom.

1) Image Preprocessing: Images tend to get de-graded as the vehicle goes underwater. Since asthe depth increases, the amount of light on ob-jects decreases and light distribution becomes non-uniform. To tackle this issue, the input imagesare converted to YCbCr color space and ContrastLimited Adaptive Histogram Equalization (CLAHE)is performed on the Luminescence(Y) channel ofthe image which improves the local contrast, atthe same time limiting the amplification whichcan over-amplify the noise too. Normalized crosscorrelation techniques and Pyramidal Mean Shiftfilter is also used to enhance the target image beforefurther processing.

Delhi Technological University Autonomous Underwater Vehicle

(a)Fig. 14. ROBOSUB Mission Controller’s Approach .

The AUV is capable of doing the tasks describedbelow. Fig.15 gives the generalized flow chart fortarget detection using vision module.

D. Gate ValidationStarting with the Gate Detection, our vision al-

gorithm receives video feed from the front facingcamera and gives the heading relative to the AUVas the output. To accomplish this the Cr channelfrom the image in YCbCr is taken. Over this image,Canny Edge Detection is performed and perpen-dicular Hough Lines are found. The center of thesymmetry of the resulting image thus gives thecorrect heading relative to the vehicle.

E. Path DetectionFor Path Detection video feed from the downward

facing camera is analyzed and heading relative tothe AUV is given as output. The captured image isconverted to HSV and segmented for specific color.Further, morphological operations are performed toremove any unnecessary noise. Contours are thenfound on the resulting image to find the connectedcomponents. Smaller contours are removed pertain-ing to noise. A minimum bounding rectangle is

formed over the detected contours and the angle ofthe edge having the longest length is taken as theangle of the path.

F. Buoy Detection

Fig. 18. Original Image and Processed Image for Buoy Detection

Algorithm for buoy detection is similar to pathdetection. Video feed from the front camera is takenand the captured image is converted to HSV andsegmented for specific color. Over the resultingbinary image, morphological operations are per-formed to remove any unnecessary noise. On theresulting image, the center of the largest blob isfound using central moments. The position of thecenter gives the relative heading for the AUV.

Delhi Technological University Autonomous Underwater Vehicle

(a) (b) (c)Fig. 16. Visual comparison: (a) Input image, (b) Cr channel from YCbCr space and (c) Gate detected after applying the algorithm

(a) (b)Fig. 17. Visual comparison: (a) Original image with path detected and heading calculated, (b) Corresponding binary image.

VII. UnderWater AcousticsZyra 2.0 has an acoustic navigation system which

employs a three dimensional array of hydrophonesfor data acquisition to calculate azimuthal and com-pass bearing and estimate hyperbolic location ofthe sound source in far-field approximation usingTime difference of arrivals (TDOAs). Fig.19 shows

Fig. 19. 3-D hydrophone array

a tetrahedral array of hydrophones which has beenused for the localization of the sound source. Sucha symmetric array eases calculation while makingtime difference of arrival(TDOA) calculations andhelps in reducing symmetric noise at different hy-drophones.

The localization task has been broken down intotwo parts as follows :

A) TDOA( Time Difference of Arrival Estimation)

B) Position Estimation

A. TDOA EstimationThe Generalized cross-correlation technique us-

ing phase transform (GCC-PHAT) is employed tocalculate the time difference of arrival correspond-ing to the correlation peak. Estimated TDOAs givebearing estimation for far field approximation. Themathematics of the method is discussed as belows:Let S 1(t) and S 2(t) be two signals at two differenthydrophones and ∆ be the Fast Fourier transform ofthe signals, then

X1( f ) = ∆(S 1(t))

X2( f ) = ∆(S 2(t))

GCC-PHAT is defined as:

GPHAT =X1( f )[X2( f )]∗

|X1( f )[X2( f )]∗|where * is the complex conjugate of a function

T DOA = argmax(∆−1(GPHAT ))

In environments of high levels of reverberationGCC-PHAT helps to improve robustness and accu-

Delhi Technological University Autonomous Underwater Vehicle

racy in calculating the time difference of arrival aswe can see from Fig.20 that GCC-PHAT enhancesthe peak and whitens the region around it.

Fig. 20. Typical GCC-PHAT function

B. Position Estimation

Fig. 21. Geometrical model of the localization system

We have to note a few terminologies here. X, Yand Z are my sensor positions respectively on thex,y and z cartesian axes. The point S represents mysound source. Since TDOA measurements require2 hydrophones, we will be dealing with sourcehydrophone triangles such as ∆OZS .

∠(ZOS ) = b is my bearing, we have defined

∠(ZS N) = a and ∠(ZNS ) = c.ZN has been drawnsuch that ZS = NS .

The path difference x = cτ where c is thespeed of sound and τ is the TDOA. We defineOZ=OX=OY=d.

We define (vectorially), OS = rs and ZS = rz andin a similar fashion we define rx and ry.

The basic TDOA equations are:-

||ri| − |r j|| = Xi j

Where i and j run over x ,y and z and x is mypath difference. Note that Xi j is symmetric due tothe modulus on the left hand side.

A little manipulation with the parallelogram lawof addition and TDOA equations would show us thatfor that triangle OZS we can derive the quadratic in|rs| as :

A = 4sin2(a/2)

B = ±4sin2(a/2)Xzs

C = (Xzs)2−d2

In A2x + Bx + C = 0, x is the dummyvaraible(different from Xi j) , where

x = |rs|Now clearly ON is the path difference that is

ON = Xzs.Determine ZN = f by cosine rule and finally using

sine rule in triangles OZN and NZS.

sin(a/2) =√

1− df sin2(b)

The discriminant of the quadratic gives us thecondition that :

cos(a/2)ϵ− dXzs, d

Xzsor in a better sense d

Xzs< 1.

The idea is to find |rs| from the quadratic andcalculate p = |rs|sin(b) and z- coordinate of sourceusing z(s) = |rs|cos(b).

And finally for Xxs or for Xys using the fact that,

|ry| =√|rs|2−2dy(s)+d2

And that y(s) = psinψ, ψ being measured in theXY plane from X axis we find out by using this inthe TDOA equation in triangle OYS.

sinψ = d2−Xys2±|rs|Xys

2dp

Delhi Technological University Autonomous Underwater Vehicle

The beauty of this method lies in the fact that wehaven’t used any approximations for calculating theexact positions .

Although there is the possibility of multiple rootsfrom the quadratic yet only one of the solutionswill be feasible satisfying all other conditions . Anyerror if encountered will be on part of the TDOAanalysis method, it will only be multiplied by ascalar constant while calculating the overall error.

VIII. Conclusion and Current ProgressIn this paper the development of the AUV Zyra

2.0 is presented to perform various shallow watertasks. The idea and logic behind the mechanicaldesign ,embedded and power systems design, con-trol algorithms and software platforms , image pro-cessing techniques and the geometric mathematicsinvolved in the localization process of the pingerhave been explained in detail.Zyra 2.0 is now inthe testing phase. Prior to vehicle assembly, themechanical systemswere thoroughly leak tested, andthe electrical systems were bench tested.

AcknowledgmentThe authors would like to thank the sponsors,

namely ONGC, Yamaha, Samtec, NIOT, Seabotixand National Instruments for their financial andtechnical support. Further thanks to our facultyadvisors P.B.Sharma and Prof. R.K.Sinha. We aregrateful to Mr. Anil for providing his pool fortesting. Without these individuals support,the devel-opment of the AUV would not have been possible.

References[1] Sinha, R.K. Delhi Technological. University and Ahmed,S. ,

Bahuguna,P. , Kumra,R. , Agarwal,V., Saxena,V. , Mittal,V. ;Gupta,P. , Tutwani,M. , Vehicle for Automation Research andUnderwater Navigation, 3rd ed.

[2] Ming Chen , Qiang Zhan and Sanlong Cai Control SystemDesign of an Autonomous Underwater Vehicle, 3rd ed. IEEEConference on Robotics, Automation and Mechatronics, Dec.2006.

[3] Daniel Moussette , Ashish Palooparambil and Jarred RaymondOptimization and Control Design of an Autonomous UnderwaterVehicle, 3rd ed. WORCESTER POLYTECHNIC INSTITUTEThursday, April 29, 2010.

[4] Ann Marie Polsenberg, MIT , Drew Gashler Developing anAUV Manual Remote Control System, 3rd ed. WORCESTERPOLYTECHNIC INSTITUTE Thursday, April 29, 2010.

[5] Robert Pavish Dynamic control capabilities developments of theBluefin Robotics AUV fleet, 3rd ed. Cambridge 2009

Delhi Technological University Autonomous Underwater Vehicle